Language impairment and communication difficulties are a hallmark of autism spectrum disorders. The central hypothesis underlying this proposal is that children on the autism spectrum manifest difficulties forming and accessing abstract representations and categorizing acoustic distinct but linguistically similar stimulus tokens into unitary objects. This difficulty has deep implications for their systems of word (lexical) processing and representation: attention to (irrelevant) acoustic detail prevents fast/successful activation of abstract lexical representations. This hypothesis will be probed in school-age children (6-10 years) on the autism spectrum (and age-matched, NVIQ-matched neurotypical controls), using the brain recording technology of magnetoencephalography (MEG) to gain spectral, temporal and spatial insights from paradigms probing lexical access, word difference detection and variants of priming. It is further hypothesized that--because failure to abstract/attention to extraneous impedes lexical access--reduction of extraneous detail (e.g. band-pass filtering) might improve abstraction/categorization by making different tokens of the same word easier to recognize as sharing an abstract representation. If successful in unifying disparate percepts of acoustically-different but lexically-identical stimuli into categorical representations/abstract representations, this manipulation might pave the way for future therapeutic options.